Used car platform Carvago has teamed up with the Technical University in Prague to develop an AI for determining the value of used cars.
The algorithm can determine the price of the vehicle based on its type, age and equipment. On average, the price determined by the AI differs by only two percent from the actual selling price.
Determining value through image and text recognition
First, the system evaluates the price on the basis of the often unstructured texts in the advertisements. In the process, the AI filters out information about the vehicle’s equipment, past repairs or the checkbook from the texts. However, the new version of the AI goes even further: in addition to the text-based price assessment, the AI also analyzes enclosed vehicle photos. In doing so, it looks at the condition of the vehicle’s paintwork, wheels and the respective optional extras
Every day, millions of offers for used cars are advertised throughout Europe. Often, the information needed to determine the vehicle’s value is missing. Given the volume of advertisements, it is almost impossible to gather this information manually. The newly developed AI is designed to address exactly this problem by performing a comprehensive and fast text and image analysis, on the basis of which it makes a precise estimate of the value of the respective used vehicle.
Cooperation between science and industry
The new AI was developed as part of a collaboration between the EAG Group, Carvago’s parent company, and artificial intelligence specialists from the Data Science Laboratory of the University Center for Energy Efficient Buildings (UCEEB) at the Czech Technical University in Prague.